Calibration of deep probabilistic models with decoupled bayesian neural networks
Entity
UAM. Departamento de Tecnología Electrónica y de las ComunicacionesPublisher
ElsevierDate
2020-09-24Citation
10.1016/j.neucom.2020.04.103
Neurocomputing 407 (2020): 194-205
ISSN
0925-2312 (print)DOI
10.1016/j.neucom.2020.04.103Funded by
We gratefully acknowledge the feedback provided by Emilio Granell and Enrique Vidal on an earlier manuscript. The authors thank the EU-FEDER Comunitat Valenciana 2014-2020 grant IDIFEDER/2018/025. We also acknowledge the support of NVIDIA by providing two GPU Titan XP from their grant program and Mario Parreño for providing the logits of the ADIENCE and VGGFACE2 models. Juan Maroñas is supported by grant FPI-UPV. Daniel Ramos is supported by the Spanish Ministry of Science, Innovation and Universities via grant RTI2018-098091-B-I00Project
Gobierno de España. RTI2018-098091-B-I00Editor's Version
https://doi.org/10.1016/j.neucom.2020.04.103Subjects
Bayesian modelling; Bayesian neural networks; Calibration; Image classification; TelecomunicacionesRights
© ElsevierAbstract
Deep Neural Networks (DNNs) have achieved state-of-the-art accuracy performance in many tasks. However, recent works have pointed out that the outputs provided by these models are not well-calibrated, seriously limiting their use in critical decision scenarios. In this work, we propose to use a decoupled Bayesian stage, implemented with a Bayesian Neural Network (BNN), to map the uncalibrated probabilities provided by a DNN to calibrated ones, consistently improving calibration. Our results evidence that incorporating uncertainty provides more reliable probabilistic models, a critical condition for achieving good calibration. We report a generous collection of experimental results using high-accuracy DNNs in standardized image classification benchmarks, showing the good performance, flexibility and robust behaviour of our approach with respect to several state-of-the-art calibration methods. Code for reproducibility is provided
Files in this item
Google Scholar:Maroñas Molano, Juan
-
Paredes, Roberto
-
Ramos Castro, Daniel
This item appears in the following Collection(s)
Related items
Showing items related by title, author, creator and subject.